I've a multiclass classification task at hand which I would like to pose as One vs One classification problems. I've models that are trained for each pair of classes and I would like to combine all of the models into a multiclass LIBSVM model file. I've been unable to find any help regarding this.
Things to add:
- I cannot pool together all the training data and train the svm again, as I dont have access to the original data that was used to train each of the individual pairwise classifiers. I only have the trained model files.
Any help regarding this would be much appreciated. Thanks!
[link][1 comment]